11 This document attempts to describe a few coding standards that are being used in
12 the LLVM source tree. Although no coding standards should be regarded as
13 absolute requirements to be followed in all instances, coding standards are
14 particularly important for large-scale code bases that follow a library-based
17 While this document may provide guidance for some mechanical formatting issues,
18 whitespace, or other "microscopic details", these are not fixed standards.
19 Always follow the golden rule:
23 **If you are extending, enhancing, or bug fixing already implemented code,
24 use the style that is already being used so that the source is uniform and
27 Note that some code bases (e.g. ``libc++``) have really good reasons to deviate
28 from the coding standards. In the case of ``libc++``, this is because the
29 naming and other conventions are dictated by the C++ standard. If you think
30 there is a specific good reason to deviate from the standards here, please bring
31 it up on the LLVMdev mailing list.
33 There are some conventions that are not uniformly followed in the code base
34 (e.g. the naming convention). This is because they are relatively new, and a
35 lot of code was written before they were put in place. Our long term goal is
36 for the entire codebase to follow the convention, but we explicitly *do not*
37 want patches that do large-scale reformating of existing code. On the other
38 hand, it is reasonable to rename the methods of a class if you're about to
39 change it in some other way. Just do the reformating as a separate commit from
40 the functionality change.
42 The ultimate goal of these guidelines is the increase readability and
43 maintainability of our common source base. If you have suggestions for topics to
44 be included, please mail them to `Chris <mailto:sabre@nondot.org>`_.
46 Languages, Libraries, and Standards
47 ===================================
49 Most source code in LLVM and other LLVM projects using these coding standards
50 is C++ code. There are some places where C code is used either due to
51 environment restrictions, historical restrictions, or due to third-party source
52 code imported into the tree. Generally, our preference is for standards
53 conforming, modern, and portable C++ code as the implementation language of
59 LLVM, Clang, and LLD are currently written using C++11 conforming code,
60 although we restrict ourselves to features which are available in the major
61 toolchains supported as host compilers. The LLDB project is even more
62 aggressive in the set of host compilers supported and thus uses still more
63 features. Regardless of the supported features, code is expected to (when
64 reasonable) be standard, portable, and modern C++11 code. We avoid unnecessary
65 vendor-specific extensions, etc.
70 Use the C++ standard library facilities whenever they are available for
71 a particular task. LLVM and related projects emphasize and rely on the standard
72 library facilities for as much as possible. Common support libraries providing
73 functionality missing from the standard library for which there are standard
74 interfaces or active work on adding standard interfaces will often be
75 implemented in the LLVM namespace following the expected standard interface.
77 There are some exceptions such as the standard I/O streams library which are
78 avoided. Also, there is much more detailed information on these subjects in the
79 :doc:`ProgrammersManual`.
81 Supported C++11 Language and Library Features
82 ---------------------------------------------
84 While LLVM, Clang, and LLD use C++11, not all features are available in all of
85 the toolchains which we support. The set of features supported for use in LLVM
86 is the intersection of those supported in MSVC 2012, GCC 4.7, and Clang 3.1.
87 The ultimate definition of this set is what build bots with those respective
88 toolchains accept. Don't argue with the build bots. However, we have some
89 guidance below to help you know what to expect.
91 Each toolchain provides a good reference for what it accepts:
93 * Clang: http://clang.llvm.org/cxx_status.html
94 * GCC: http://gcc.gnu.org/projects/cxx0x.html
95 * MSVC: http://msdn.microsoft.com/en-us/library/hh567368.aspx
97 In most cases, the MSVC list will be the dominating factor. Here is a summary
98 of the features that are expected to work. Features not on this list are
99 unlikely to be supported by our host compilers.
101 * Rvalue references: N2118_
103 * But *not* Rvalue references for ``*this`` or member qualifiers (N2439_)
105 * Static assert: N1720_
106 * ``auto`` type deduction: N1984_, N1737_
107 * Trailing return types: N2541_
110 * But *not* ``std::function``, until Clang implements `MSVC-compatible RTTI`_.
111 * And *not* lambdas with default arguments.
113 * ``decltype``: N2343_
114 * Nested closing right angle brackets: N1757_
115 * Extern templates: N1987_
116 * ``nullptr``: N2431_
117 * Strongly-typed and forward declarable enums: N2347_, N2764_
118 * Local and unnamed types as template arguments: N2657_
119 * Range-based for-loop: N2930_
121 * But ``{}`` are required around inner ``do {} while()`` loops. As a result,
122 ``{}`` are required around function-like macros inside range-based for
125 * ``override`` and ``final``: N2928_, N3206_, N3272_
126 * Atomic operations and the C++11 memory model: N2429_
128 .. _N2118: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n2118.html
129 .. _N2439: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2439.htm
130 .. _N1720: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2004/n1720.html
131 .. _N1984: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n1984.pdf
132 .. _N1737: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2004/n1737.pdf
133 .. _N2541: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2541.htm
134 .. _N2927: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2009/n2927.pdf
135 .. _N2343: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2343.pdf
136 .. _N1757: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2005/n1757.html
137 .. _N1987: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n1987.htm
138 .. _N2431: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2431.pdf
139 .. _N2347: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2347.pdf
140 .. _N2764: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2764.pdf
141 .. _N2657: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2657.htm
142 .. _N2930: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2009/n2930.html
143 .. _N2928: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2009/n2928.htm
144 .. _N3206: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2010/n3206.htm
145 .. _N3272: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3272.htm
146 .. _N2429: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2429.htm
147 .. _MSVC-compatible RTTI: http://llvm.org/PR18951
149 The supported features in the C++11 standard libraries are less well tracked,
150 but also much greater. Most of the standard libraries implement most of C++11's
151 library. The most likely lowest common denominator is Linux support. For
152 libc++, the support is just poorly tested and undocumented but expected to be
153 largely complete. YMMV. For libstdc++, the support is documented in detail in
154 `the libstdc++ manual`_. There are some very minor missing facilities that are
155 unlikely to be common problems, and there are a few larger gaps that are worth
158 * Not all of the type traits are implemented
159 * No regular expression library.
160 * While most of the atomics library is well implemented, the fences are
161 missing. Fortunately, they are rarely needed.
162 * The locale support is incomplete.
163 * ``std::initializer_list`` (and the constructors and functions that take it as
164 an argument) are not always available, so you cannot (for example) initialize
165 a ``std::vector`` with a braced initializer list.
167 Other than these areas you should assume the standard library is available and
168 working as expected until some build bot tells you otherwise. If you're in an
169 uncertain area of one of the above points, but you cannot test on a Linux
170 system, your best approach is to minimize your use of these features, and watch
171 the Linux build bots to find out if your usage triggered a bug. For example, if
172 you hit a type trait which doesn't work we can then add support to LLVM's
173 traits header to emulate it.
175 .. _the libstdc++ manual:
176 http://gcc.gnu.org/onlinedocs/gcc-4.7.3/libstdc++/manual/manual/status.html#status.iso.2011
178 Mechanical Source Issues
179 ========================
181 Source Code Formatting
182 ----------------------
187 Comments are one critical part of readability and maintainability. Everyone
188 knows they should comment their code, and so should you. When writing comments,
189 write them as English prose, which means they should use proper capitalization,
190 punctuation, etc. Aim to describe what the code is trying to do and why, not
191 *how* it does it at a micro level. Here are a few critical things to document:
193 .. _header file comment:
198 Every source file should have a header on it that describes the basic purpose of
199 the file. If a file does not have a header, it should not be checked into the
200 tree. The standard header looks like this:
204 //===-- llvm/Instruction.h - Instruction class definition -------*- C++ -*-===//
206 // The LLVM Compiler Infrastructure
208 // This file is distributed under the University of Illinois Open Source
209 // License. See LICENSE.TXT for details.
211 //===----------------------------------------------------------------------===//
214 /// \brief This file contains the declaration of the Instruction class, which is
215 /// the base class for all of the VM instructions.
217 //===----------------------------------------------------------------------===//
219 A few things to note about this particular format: The "``-*- C++ -*-``" string
220 on the first line is there to tell Emacs that the source file is a C++ file, not
221 a C file (Emacs assumes ``.h`` files are C files by default).
225 This tag is not necessary in ``.cpp`` files. The name of the file is also
226 on the first line, along with a very short description of the purpose of the
227 file. This is important when printing out code and flipping though lots of
230 The next section in the file is a concise note that defines the license that the
231 file is released under. This makes it perfectly clear what terms the source
232 code can be distributed under and should not be modified in any way.
234 The main body is a ``doxygen`` comment describing the purpose of the file. It
235 should have a ``\brief`` command that describes the file in one or two
236 sentences. Any additional information should be separated by a blank line. If
237 an algorithm is being implemented or something tricky is going on, a reference
238 to the paper where it is published should be included, as well as any notes or
239 *gotchas* in the code to watch out for.
244 Classes are one fundamental part of a good object oriented design. As such, a
245 class definition should have a comment block that explains what the class is
246 used for and how it works. Every non-trivial class is expected to have a
247 ``doxygen`` comment block.
252 Methods defined in a class (as well as any global functions) should also be
253 documented properly. A quick note about what it does and a description of the
254 borderline behaviour is all that is necessary here (unless something
255 particularly tricky or insidious is going on). The hope is that people can
256 figure out how to use your interfaces without reading the code itself.
258 Good things to talk about here are what happens when something unexpected
259 happens: does the method return null? Abort? Format your hard disk?
264 In general, prefer C++ style (``//``) comments. They take less space, require
265 less typing, don't have nesting problems, etc. There are a few cases when it is
266 useful to use C style (``/* */``) comments however:
268 #. When writing C code: Obviously if you are writing C code, use C style
271 #. When writing a header file that may be ``#include``\d by a C source file.
273 #. When writing a source file that is used by a tool that only accepts C style
276 To comment out a large block of code, use ``#if 0`` and ``#endif``. These nest
277 properly and are better behaved in general than C style comments.
279 Doxygen Use in Documentation Comments
280 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
282 Use the ``\file`` command to turn the standard file header into a file-level
285 Include descriptive ``\brief`` paragraphs for all public interfaces (public
286 classes, member and non-member functions). Explain API use and purpose in
287 ``\brief`` paragraphs, don't just restate the information that can be inferred
288 from the API name. Put detailed discussion into separate paragraphs.
290 To refer to parameter names inside a paragraph, use the ``\p name`` command.
291 Don't use the ``\arg name`` command since it starts a new paragraph that
292 contains documentation for the parameter.
294 Wrap non-inline code examples in ``\code ... \endcode``.
296 To document a function parameter, start a new paragraph with the
297 ``\param name`` command. If the parameter is used as an out or an in/out
298 parameter, use the ``\param [out] name`` or ``\param [in,out] name`` command,
301 To describe function return value, start a new paragraph with the ``\returns``
304 A minimal documentation comment:
308 /// \brief Does foo and bar.
309 void fooBar(bool Baz);
311 A documentation comment that uses all Doxygen features in a preferred way:
315 /// \brief Does foo and bar.
317 /// Does not do foo the usual way if \p Baz is true.
321 /// fooBar(false, "quux", Res);
324 /// \param Quux kind of foo to do.
325 /// \param [out] Result filled with bar sequence on foo success.
327 /// \returns true on success.
328 bool fooBar(bool Baz, StringRef Quux, std::vector<int> &Result);
330 Don't duplicate the documentation comment in the header file and in the
331 implementation file. Put the documentation comments for public APIs into the
332 header file. Documentation comments for private APIs can go to the
333 implementation file. In any case, implementation files can include additional
334 comments (not necessarily in Doxygen markup) to explain implementation details
337 Don't duplicate function or class name at the beginning of the comment.
338 For humans it is obvious which function or class is being documented;
339 automatic documentation processing tools are smart enough to bind the comment
340 to the correct declaration.
348 /// Something - An abstraction for some complicated thing.
351 /// fooBar - Does foo and bar.
357 /// fooBar - Does foo and bar.
358 void Something::fooBar() { ... }
366 /// \brief An abstraction for some complicated thing.
369 /// \brief Does foo and bar.
375 // Builds a B-tree in order to do foo. See paper by...
376 void Something::fooBar() { ... }
378 It is not required to use additional Doxygen features, but sometimes it might
379 be a good idea to do so.
383 * adding comments to any narrow namespace containing a collection of
384 related functions or types;
386 * using top-level groups to organize a collection of related functions at
387 namespace scope where the grouping is smaller than the namespace;
389 * using member groups and additional comments attached to member
390 groups to organize within a class.
397 /// \name Functions that do Foo.
408 Immediately after the `header file comment`_ (and include guards if working on a
409 header file), the `minimal list of #includes`_ required by the file should be
410 listed. We prefer these ``#include``\s to be listed in this order:
412 .. _Main Module Header:
413 .. _Local/Private Headers:
415 #. Main Module Header
416 #. Local/Private Headers
418 #. System ``#include``\s
420 and each category should be sorted lexicographically by the full path.
422 The `Main Module Header`_ file applies to ``.cpp`` files which implement an
423 interface defined by a ``.h`` file. This ``#include`` should always be included
424 **first** regardless of where it lives on the file system. By including a
425 header file first in the ``.cpp`` files that implement the interfaces, we ensure
426 that the header does not have any hidden dependencies which are not explicitly
427 ``#include``\d in the header, but should be. It is also a form of documentation
428 in the ``.cpp`` file to indicate where the interfaces it implements are defined.
430 .. _fit into 80 columns:
435 Write your code to fit within 80 columns of text. This helps those of us who
436 like to print out code and look at your code in an ``xterm`` without resizing
439 The longer answer is that there must be some limit to the width of the code in
440 order to reasonably allow developers to have multiple files side-by-side in
441 windows on a modest display. If you are going to pick a width limit, it is
442 somewhat arbitrary but you might as well pick something standard. Going with 90
443 columns (for example) instead of 80 columns wouldn't add any significant value
444 and would be detrimental to printing out code. Also many other projects have
445 standardized on 80 columns, so some people have already configured their editors
446 for it (vs something else, like 90 columns).
448 This is one of many contentious issues in coding standards, but it is not up for
451 Use Spaces Instead of Tabs
452 ^^^^^^^^^^^^^^^^^^^^^^^^^^
454 In all cases, prefer spaces to tabs in source files. People have different
455 preferred indentation levels, and different styles of indentation that they
456 like; this is fine. What isn't fine is that different editors/viewers expand
457 tabs out to different tab stops. This can cause your code to look completely
458 unreadable, and it is not worth dealing with.
460 As always, follow the `Golden Rule`_ above: follow the style of
461 existing code if you are modifying and extending it. If you like four spaces of
462 indentation, **DO NOT** do that in the middle of a chunk of code with two spaces
463 of indentation. Also, do not reindent a whole source file: it makes for
464 incredible diffs that are absolutely worthless.
466 Indent Code Consistently
467 ^^^^^^^^^^^^^^^^^^^^^^^^
469 Okay, in your first year of programming you were told that indentation is
470 important. If you didn't believe and internalize this then, now is the time.
471 Just do it. With the introduction of C++11, there are some new formatting
472 challenges that merit some suggestions to help have consistent, maintainable,
473 and tool-friendly formatting and indentation.
475 Format Lambdas Like Blocks Of Code
476 """"""""""""""""""""""""""""""""""
478 When formatting a multi-line lambda, format it like a block of code, that's
479 what it is. If there is only one multi-line lambda in a statement, and there
480 are no expressions lexically after it in the statement, drop the indent to the
481 standard two space indent for a block of code, as if it were an if-block opened
482 by the preceding part of the statement:
486 std::sort(foo.begin(), foo.end(), [&](Foo a, Foo b) -> bool {
491 return a.bam < b.bam;
494 To take best advantage of this formatting, if you are designing an API which
495 accepts a continuation or single callable argument (be it a functor, or
496 a ``std::function``), it should be the last argument if at all possible.
498 If there are multiple multi-line lambdas in a statement, or there is anything
499 interesting after the lambda in the statement, indent the block two spaces from
500 the indent of the ``[]``:
504 dyn_switch(V->stripPointerCasts(),
508 [] (SelectInst *SI) {
509 // process selects...
514 [] (AllocaInst *AI) {
515 // process allocas...
518 Braced Initializer Lists
519 """"""""""""""""""""""""
521 With C++11, there are significantly more uses of braced lists to perform
522 initialization. These allow you to easily construct aggregate temporaries in
523 expressions among other niceness. They now have a natural way of ending up
524 nested within each other and within function calls in order to build up
525 aggregates (such as option structs) from local variables. To make matters
526 worse, we also have many more uses of braces in an expression context that are
527 *not* performing initialization.
529 The historically common formatting of braced initialization of aggregate
530 variables does not mix cleanly with deep nesting, general expression contexts,
531 function arguments, and lambdas. We suggest new code use a simple rule for
532 formatting braced initialization lists: act as-if the braces were parentheses
533 in a function call. The formatting rules exactly match those already well
534 understood for formatting nested function calls. Examples:
538 foo({a, b, c}, {1, 2, 3});
540 llvm::Constant *Mask[] = {
541 llvm::ConstantInt::get(llvm::Type::getInt32Ty(getLLVMContext()), 0),
542 llvm::ConstantInt::get(llvm::Type::getInt32Ty(getLLVMContext()), 1),
543 llvm::ConstantInt::get(llvm::Type::getInt32Ty(getLLVMContext()), 2)};
545 This formatting scheme also makes it particularly easy to get predictable,
546 consistent, and automatic formatting with tools like `Clang Format`_.
548 .. _Clang Format: http://clang.llvm.org/docs/ClangFormat.html
550 Language and Compiler Issues
551 ----------------------------
553 Treat Compiler Warnings Like Errors
554 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
556 If your code has compiler warnings in it, something is wrong --- you aren't
557 casting values correctly, you have "questionable" constructs in your code, or
558 you are doing something legitimately wrong. Compiler warnings can cover up
559 legitimate errors in output and make dealing with a translation unit difficult.
561 It is not possible to prevent all warnings from all compilers, nor is it
562 desirable. Instead, pick a standard compiler (like ``gcc``) that provides a
563 good thorough set of warnings, and stick to it. At least in the case of
564 ``gcc``, it is possible to work around any spurious errors by changing the
565 syntax of the code slightly. For example, a warning that annoys me occurs when
566 I write code like this:
570 if (V = getValue()) {
574 ``gcc`` will warn me that I probably want to use the ``==`` operator, and that I
575 probably mistyped it. In most cases, I haven't, and I really don't want the
576 spurious errors. To fix this particular problem, I rewrite the code like
581 if ((V = getValue())) {
585 which shuts ``gcc`` up. Any ``gcc`` warning that annoys you can be fixed by
586 massaging the code appropriately.
591 In almost all cases, it is possible and within reason to write completely
592 portable code. If there are cases where it isn't possible to write portable
593 code, isolate it behind a well defined (and well documented) interface.
595 In practice, this means that you shouldn't assume much about the host compiler
596 (and Visual Studio tends to be the lowest common denominator). If advanced
597 features are used, they should only be an implementation detail of a library
598 which has a simple exposed API, and preferably be buried in ``libSystem``.
600 Do not use RTTI or Exceptions
601 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
603 In an effort to reduce code and executable size, LLVM does not use RTTI
604 (e.g. ``dynamic_cast<>;``) or exceptions. These two language features violate
605 the general C++ principle of *"you only pay for what you use"*, causing
606 executable bloat even if exceptions are never used in the code base, or if RTTI
607 is never used for a class. Because of this, we turn them off globally in the
610 That said, LLVM does make extensive use of a hand-rolled form of RTTI that use
611 templates like :ref:`isa\<>, cast\<>, and dyn_cast\<> <isa>`.
612 This form of RTTI is opt-in and can be
613 :doc:`added to any class <HowToSetUpLLVMStyleRTTI>`. It is also
614 substantially more efficient than ``dynamic_cast<>``.
616 .. _static constructor:
618 Do not use Static Constructors
619 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
621 Static constructors and destructors (e.g. global variables whose types have a
622 constructor or destructor) should not be added to the code base, and should be
623 removed wherever possible. Besides `well known problems
624 <http://yosefk.com/c++fqa/ctors.html#fqa-10.12>`_ where the order of
625 initialization is undefined between globals in different source files, the
626 entire concept of static constructors is at odds with the common use case of
627 LLVM as a library linked into a larger application.
629 Consider the use of LLVM as a JIT linked into another application (perhaps for
630 `OpenGL, custom languages <http://llvm.org/Users.html>`_, `shaders in movies
631 <http://llvm.org/devmtg/2010-11/Gritz-OpenShadingLang.pdf>`_, etc). Due to the
632 design of static constructors, they must be executed at startup time of the
633 entire application, regardless of whether or how LLVM is used in that larger
634 application. There are two problems with this:
636 * The time to run the static constructors impacts startup time of applications
637 --- a critical time for GUI apps, among others.
639 * The static constructors cause the app to pull many extra pages of memory off
640 the disk: both the code for the constructor in each ``.o`` file and the small
641 amount of data that gets touched. In addition, touched/dirty pages put more
642 pressure on the VM system on low-memory machines.
644 We would really like for there to be zero cost for linking in an additional LLVM
645 target or other library into an application, but static constructors violate
648 That said, LLVM unfortunately does contain static constructors. It would be a
649 `great project <http://llvm.org/PR11944>`_ for someone to purge all static
650 constructors from LLVM, and then enable the ``-Wglobal-constructors`` warning
651 flag (when building with Clang) to ensure we do not regress in the future.
653 Use of ``class`` and ``struct`` Keywords
654 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
656 In C++, the ``class`` and ``struct`` keywords can be used almost
657 interchangeably. The only difference is when they are used to declare a class:
658 ``class`` makes all members private by default while ``struct`` makes all
659 members public by default.
661 Unfortunately, not all compilers follow the rules and some will generate
662 different symbols based on whether ``class`` or ``struct`` was used to declare
663 the symbol (e.g., MSVC). This can lead to problems at link time.
665 * All declarations and definitions of a given ``class`` or ``struct`` must use
666 the same keyword. For example:
672 // Breaks mangling in MSVC.
673 struct Foo { int Data; };
675 * As a rule of thumb, ``struct`` should be kept to structures where *all*
676 members are declared public.
680 // Foo feels like a class... this is strange.
686 int getData() const { return Data; }
687 void setData(int D) { Data = D; }
690 // Bar isn't POD, but it does look like a struct.
696 Do not use Braced Initializer Lists to Call a Constructor
697 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
699 In C++11 there is a "generalized initialization syntax" which allows calling
700 constructors using braced initializer lists. Do not use these to call
701 constructors with any interesting logic or if you care that you're calling some
702 *particular* constructor. Those should look like function calls using
703 parentheses rather than like aggregate initialization. Similarly, if you need
704 to explicitly name the type and call its constructor to create a temporary,
705 don't use a braced initializer list. Instead, use a braced initializer list
706 (without any type for temporaries) when doing aggregate initialization or
707 something notionally equivalent. Examples:
713 // Construct a Foo by reading data from the disk in the whizbang format, ...
714 Foo(std::string filename);
716 // Construct a Foo by looking up the Nth element of some global data ...
722 // The Foo constructor call is very deliberate, no braces.
723 std::fill(foo.begin(), foo.end(), Foo("name"));
725 // The pair is just being constructed like an aggregate, use braces.
726 bar_map.insert({my_key, my_value});
728 If you use a braced initializer list when initializing a variable, use an equals before the open curly brace:
732 int data[] = {0, 1, 2, 3};
734 Use ``auto`` Type Deduction to Make Code More Readable
735 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
737 Some are advocating a policy of "almost always ``auto``" in C++11, however LLVM
738 uses a more moderate stance. Use ``auto`` if and only if it makes the code more
739 readable or easier to maintain. Don't "almost always" use ``auto``, but do use
740 ``auto`` with initializers like ``cast<Foo>(...)`` or other places where the
741 type is already obvious from the context. Another time when ``auto`` works well
742 for these purposes is when the type would have been abstracted away anyways,
743 often behind a container's typedef such as ``std::vector<T>::iterator``.
745 Beware unnecessary copies with ``auto``
746 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
748 The convenience of ``auto`` makes it easy to forget that its default behavior
749 is a copy. Particularly in range-based ``for`` loops, careless copies are
752 As a rule of thumb, use ``auto &`` unless you need to copy the result, and use
753 ``auto *`` when copying pointers.
757 // Typically there's no reason to copy.
758 for (const auto &Val : Container) { observe(Val); }
759 for (auto &Val : Container) { Val.change(); }
761 // Remove the reference if you really want a new copy.
762 for (auto Val : Container) { Val.change(); saveSomewhere(Val); }
764 // Copy pointers, but make it clear that they're pointers.
765 for (const auto *Ptr : Container) { observe(*Ptr); }
766 for (auto *Ptr : Container) { Ptr->change(); }
771 The High-Level Issues
772 ---------------------
774 A Public Header File **is** a Module
775 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
777 C++ doesn't do too well in the modularity department. There is no real
778 encapsulation or data hiding (unless you use expensive protocol classes), but it
779 is what we have to work with. When you write a public header file (in the LLVM
780 source tree, they live in the top level "``include``" directory), you are
781 defining a module of functionality.
783 Ideally, modules should be completely independent of each other, and their
784 header files should only ``#include`` the absolute minimum number of headers
785 possible. A module is not just a class, a function, or a namespace: it's a
786 collection of these that defines an interface. This interface may be several
787 functions, classes, or data structures, but the important issue is how they work
790 In general, a module should be implemented by one or more ``.cpp`` files. Each
791 of these ``.cpp`` files should include the header that defines their interface
792 first. This ensures that all of the dependences of the module header have been
793 properly added to the module header itself, and are not implicit. System
794 headers should be included after user headers for a translation unit.
796 .. _minimal list of #includes:
798 ``#include`` as Little as Possible
799 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
801 ``#include`` hurts compile time performance. Don't do it unless you have to,
802 especially in header files.
804 But wait! Sometimes you need to have the definition of a class to use it, or to
805 inherit from it. In these cases go ahead and ``#include`` that header file. Be
806 aware however that there are many cases where you don't need to have the full
807 definition of a class. If you are using a pointer or reference to a class, you
808 don't need the header file. If you are simply returning a class instance from a
809 prototyped function or method, you don't need it. In fact, for most cases, you
810 simply don't need the definition of a class. And not ``#include``\ing speeds up
813 It is easy to try to go too overboard on this recommendation, however. You
814 **must** include all of the header files that you are using --- you can include
815 them either directly or indirectly through another header file. To make sure
816 that you don't accidentally forget to include a header file in your module
817 header, make sure to include your module header **first** in the implementation
818 file (as mentioned above). This way there won't be any hidden dependencies that
819 you'll find out about later.
821 Keep "Internal" Headers Private
822 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
824 Many modules have a complex implementation that causes them to use more than one
825 implementation (``.cpp``) file. It is often tempting to put the internal
826 communication interface (helper classes, extra functions, etc) in the public
827 module header file. Don't do this!
829 If you really need to do something like this, put a private header file in the
830 same directory as the source files, and include it locally. This ensures that
831 your private interface remains private and undisturbed by outsiders.
835 It's okay to put extra implementation methods in a public class itself. Just
836 make them private (or protected) and all is well.
840 Use Early Exits and ``continue`` to Simplify Code
841 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
843 When reading code, keep in mind how much state and how many previous decisions
844 have to be remembered by the reader to understand a block of code. Aim to
845 reduce indentation where possible when it doesn't make it more difficult to
846 understand the code. One great way to do this is by making use of early exits
847 and the ``continue`` keyword in long loops. As an example of using an early
848 exit from a function, consider this "bad" code:
852 Value *doSomething(Instruction *I) {
853 if (!isa<TerminatorInst>(I) &&
854 I->hasOneUse() && doOtherThing(I)) {
855 ... some long code ....
861 This code has several problems if the body of the ``'if'`` is large. When
862 you're looking at the top of the function, it isn't immediately clear that this
863 *only* does interesting things with non-terminator instructions, and only
864 applies to things with the other predicates. Second, it is relatively difficult
865 to describe (in comments) why these predicates are important because the ``if``
866 statement makes it difficult to lay out the comments. Third, when you're deep
867 within the body of the code, it is indented an extra level. Finally, when
868 reading the top of the function, it isn't clear what the result is if the
869 predicate isn't true; you have to read to the end of the function to know that
872 It is much preferred to format the code like this:
876 Value *doSomething(Instruction *I) {
877 // Terminators never need 'something' done to them because ...
878 if (isa<TerminatorInst>(I))
881 // We conservatively avoid transforming instructions with multiple uses
882 // because goats like cheese.
886 // This is really just here for example.
887 if (!doOtherThing(I))
890 ... some long code ....
893 This fixes these problems. A similar problem frequently happens in ``for``
894 loops. A silly example is something like this:
898 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
899 if (BinaryOperator *BO = dyn_cast<BinaryOperator>(II)) {
900 Value *LHS = BO->getOperand(0);
901 Value *RHS = BO->getOperand(1);
908 When you have very, very small loops, this sort of structure is fine. But if it
909 exceeds more than 10-15 lines, it becomes difficult for people to read and
910 understand at a glance. The problem with this sort of code is that it gets very
911 nested very quickly. Meaning that the reader of the code has to keep a lot of
912 context in their brain to remember what is going immediately on in the loop,
913 because they don't know if/when the ``if`` conditions will have ``else``\s etc.
914 It is strongly preferred to structure the loop like this:
918 for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E; ++II) {
919 BinaryOperator *BO = dyn_cast<BinaryOperator>(II);
922 Value *LHS = BO->getOperand(0);
923 Value *RHS = BO->getOperand(1);
924 if (LHS == RHS) continue;
929 This has all the benefits of using early exits for functions: it reduces nesting
930 of the loop, it makes it easier to describe why the conditions are true, and it
931 makes it obvious to the reader that there is no ``else`` coming up that they
932 have to push context into their brain for. If a loop is large, this can be a
933 big understandability win.
935 Don't use ``else`` after a ``return``
936 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
938 For similar reasons above (reduction of indentation and easier reading), please
939 do not use ``'else'`` or ``'else if'`` after something that interrupts control
940 flow --- like ``return``, ``break``, ``continue``, ``goto``, etc. For
941 example, this is *bad*:
947 Type = Context.getsigjmp_bufType();
949 Error = ASTContext::GE_Missing_sigjmp_buf;
955 Type = Context.getjmp_bufType();
957 Error = ASTContext::GE_Missing_jmp_buf;
965 It is better to write it like this:
971 Type = Context.getsigjmp_bufType();
973 Error = ASTContext::GE_Missing_sigjmp_buf;
977 Type = Context.getjmp_bufType();
979 Error = ASTContext::GE_Missing_jmp_buf;
985 Or better yet (in this case) as:
991 Type = Context.getsigjmp_bufType();
993 Type = Context.getjmp_bufType();
996 Error = Signed ? ASTContext::GE_Missing_sigjmp_buf :
997 ASTContext::GE_Missing_jmp_buf;
1002 The idea is to reduce indentation and the amount of code you have to keep track
1003 of when reading the code.
1005 Turn Predicate Loops into Predicate Functions
1006 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1008 It is very common to write small loops that just compute a boolean value. There
1009 are a number of ways that people commonly write these, but an example of this
1014 bool FoundFoo = false;
1015 for (unsigned I = 0, E = BarList.size(); I != E; ++I)
1016 if (BarList[I]->isFoo()) {
1025 This sort of code is awkward to write, and is almost always a bad sign. Instead
1026 of this sort of loop, we strongly prefer to use a predicate function (which may
1027 be `static`_) that uses `early exits`_ to compute the predicate. We prefer the
1028 code to be structured like this:
1032 /// \returns true if the specified list has an element that is a foo.
1033 static bool containsFoo(const std::vector<Bar*> &List) {
1034 for (unsigned I = 0, E = List.size(); I != E; ++I)
1035 if (List[I]->isFoo())
1041 if (containsFoo(BarList)) {
1045 There are many reasons for doing this: it reduces indentation and factors out
1046 code which can often be shared by other code that checks for the same predicate.
1047 More importantly, it *forces you to pick a name* for the function, and forces
1048 you to write a comment for it. In this silly example, this doesn't add much
1049 value. However, if the condition is complex, this can make it a lot easier for
1050 the reader to understand the code that queries for this predicate. Instead of
1051 being faced with the in-line details of how we check to see if the BarList
1052 contains a foo, we can trust the function name and continue reading with better
1055 The Low-Level Issues
1056 --------------------
1058 Name Types, Functions, Variables, and Enumerators Properly
1059 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1061 Poorly-chosen names can mislead the reader and cause bugs. We cannot stress
1062 enough how important it is to use *descriptive* names. Pick names that match
1063 the semantics and role of the underlying entities, within reason. Avoid
1064 abbreviations unless they are well known. After picking a good name, make sure
1065 to use consistent capitalization for the name, as inconsistency requires clients
1066 to either memorize the APIs or to look it up to find the exact spelling.
1068 In general, names should be in camel case (e.g. ``TextFileReader`` and
1069 ``isLValue()``). Different kinds of declarations have different rules:
1071 * **Type names** (including classes, structs, enums, typedefs, etc) should be
1072 nouns and start with an upper-case letter (e.g. ``TextFileReader``).
1074 * **Variable names** should be nouns (as they represent state). The name should
1075 be camel case, and start with an upper case letter (e.g. ``Leader`` or
1078 * **Function names** should be verb phrases (as they represent actions), and
1079 command-like function should be imperative. The name should be camel case,
1080 and start with a lower case letter (e.g. ``openFile()`` or ``isFoo()``).
1082 * **Enum declarations** (e.g. ``enum Foo {...}``) are types, so they should
1083 follow the naming conventions for types. A common use for enums is as a
1084 discriminator for a union, or an indicator of a subclass. When an enum is
1085 used for something like this, it should have a ``Kind`` suffix
1086 (e.g. ``ValueKind``).
1088 * **Enumerators** (e.g. ``enum { Foo, Bar }``) and **public member variables**
1089 should start with an upper-case letter, just like types. Unless the
1090 enumerators are defined in their own small namespace or inside a class,
1091 enumerators should have a prefix corresponding to the enum declaration name.
1092 For example, ``enum ValueKind { ... };`` may contain enumerators like
1093 ``VK_Argument``, ``VK_BasicBlock``, etc. Enumerators that are just
1094 convenience constants are exempt from the requirement for a prefix. For
1104 As an exception, classes that mimic STL classes can have member names in STL's
1105 style of lower-case words separated by underscores (e.g. ``begin()``,
1106 ``push_back()``, and ``empty()``). Classes that provide multiple
1107 iterators should add a singular prefix to ``begin()`` and ``end()``
1108 (e.g. ``global_begin()`` and ``use_begin()``).
1110 Here are some examples of good and bad names:
1114 class VehicleMaker {
1116 Factory<Tire> F; // Bad -- abbreviation and non-descriptive.
1117 Factory<Tire> Factory; // Better.
1118 Factory<Tire> TireFactory; // Even better -- if VehicleMaker has more than one
1119 // kind of factories.
1122 Vehicle MakeVehicle(VehicleType Type) {
1123 VehicleMaker M; // Might be OK if having a short life-span.
1124 Tire Tmp1 = M.makeTire(); // Bad -- 'Tmp1' provides no information.
1125 Light Headlight = M.makeLight("head"); // Good -- descriptive.
1132 Use the "``assert``" macro to its fullest. Check all of your preconditions and
1133 assumptions, you never know when a bug (not necessarily even yours) might be
1134 caught early by an assertion, which reduces debugging time dramatically. The
1135 "``<cassert>``" header file is probably already included by the header files you
1136 are using, so it doesn't cost anything to use it.
1138 To further assist with debugging, make sure to put some kind of error message in
1139 the assertion statement, which is printed if the assertion is tripped. This
1140 helps the poor debugger make sense of why an assertion is being made and
1141 enforced, and hopefully what to do about it. Here is one complete example:
1145 inline Value *getOperand(unsigned I) {
1146 assert(I < Operands.size() && "getOperand() out of range!");
1150 Here are more examples:
1154 assert(Ty->isPointerType() && "Can't allocate a non-pointer type!");
1156 assert((Opcode == Shl || Opcode == Shr) && "ShiftInst Opcode invalid!");
1158 assert(idx < getNumSuccessors() && "Successor # out of range!");
1160 assert(V1.getType() == V2.getType() && "Constant types must be identical!");
1162 assert(isa<PHINode>(Succ->front()) && "Only works on PHId BBs!");
1166 In the past, asserts were used to indicate a piece of code that should not be
1167 reached. These were typically of the form:
1171 assert(0 && "Invalid radix for integer literal");
1173 This has a few issues, the main one being that some compilers might not
1174 understand the assertion, or warn about a missing return in builds where
1175 assertions are compiled out.
1177 Today, we have something much better: ``llvm_unreachable``:
1181 llvm_unreachable("Invalid radix for integer literal");
1183 When assertions are enabled, this will print the message if it's ever reached
1184 and then exit the program. When assertions are disabled (i.e. in release
1185 builds), ``llvm_unreachable`` becomes a hint to compilers to skip generating
1186 code for this branch. If the compiler does not support this, it will fall back
1187 to the "abort" implementation.
1189 Another issue is that values used only by assertions will produce an "unused
1190 value" warning when assertions are disabled. For example, this code will warn:
1194 unsigned Size = V.size();
1195 assert(Size > 42 && "Vector smaller than it should be");
1197 bool NewToSet = Myset.insert(Value);
1198 assert(NewToSet && "The value shouldn't be in the set yet");
1200 These are two interesting different cases. In the first case, the call to
1201 ``V.size()`` is only useful for the assert, and we don't want it executed when
1202 assertions are disabled. Code like this should move the call into the assert
1203 itself. In the second case, the side effects of the call must happen whether
1204 the assert is enabled or not. In this case, the value should be cast to void to
1205 disable the warning. To be specific, it is preferred to write the code like
1210 assert(V.size() > 42 && "Vector smaller than it should be");
1212 bool NewToSet = Myset.insert(Value); (void)NewToSet;
1213 assert(NewToSet && "The value shouldn't be in the set yet");
1215 Do Not Use ``using namespace std``
1216 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1218 In LLVM, we prefer to explicitly prefix all identifiers from the standard
1219 namespace with an "``std::``" prefix, rather than rely on "``using namespace
1222 In header files, adding a ``'using namespace XXX'`` directive pollutes the
1223 namespace of any source file that ``#include``\s the header. This is clearly a
1226 In implementation files (e.g. ``.cpp`` files), the rule is more of a stylistic
1227 rule, but is still important. Basically, using explicit namespace prefixes
1228 makes the code **clearer**, because it is immediately obvious what facilities
1229 are being used and where they are coming from. And **more portable**, because
1230 namespace clashes cannot occur between LLVM code and other namespaces. The
1231 portability rule is important because different standard library implementations
1232 expose different symbols (potentially ones they shouldn't), and future revisions
1233 to the C++ standard will add more symbols to the ``std`` namespace. As such, we
1234 never use ``'using namespace std;'`` in LLVM.
1236 The exception to the general rule (i.e. it's not an exception for the ``std``
1237 namespace) is for implementation files. For example, all of the code in the
1238 LLVM project implements code that lives in the 'llvm' namespace. As such, it is
1239 ok, and actually clearer, for the ``.cpp`` files to have a ``'using namespace
1240 llvm;'`` directive at the top, after the ``#include``\s. This reduces
1241 indentation in the body of the file for source editors that indent based on
1242 braces, and keeps the conceptual context cleaner. The general form of this rule
1243 is that any ``.cpp`` file that implements code in any namespace may use that
1244 namespace (and its parents'), but should not use any others.
1246 Provide a Virtual Method Anchor for Classes in Headers
1247 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1249 If a class is defined in a header file and has a vtable (either it has virtual
1250 methods or it derives from classes with virtual methods), it must always have at
1251 least one out-of-line virtual method in the class. Without this, the compiler
1252 will copy the vtable and RTTI into every ``.o`` file that ``#include``\s the
1253 header, bloating ``.o`` file sizes and increasing link times.
1255 Don't use default labels in fully covered switches over enumerations
1256 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1258 ``-Wswitch`` warns if a switch, without a default label, over an enumeration
1259 does not cover every enumeration value. If you write a default label on a fully
1260 covered switch over an enumeration then the ``-Wswitch`` warning won't fire
1261 when new elements are added to that enumeration. To help avoid adding these
1262 kinds of defaults, Clang has the warning ``-Wcovered-switch-default`` which is
1263 off by default but turned on when building LLVM with a version of Clang that
1264 supports the warning.
1266 A knock-on effect of this stylistic requirement is that when building LLVM with
1267 GCC you may get warnings related to "control may reach end of non-void function"
1268 if you return from each case of a covered switch-over-enum because GCC assumes
1269 that the enum expression may take any representable value, not just those of
1270 individual enumerators. To suppress this warning, use ``llvm_unreachable`` after
1273 Use ``LLVM_DELETED_FUNCTION`` to mark uncallable methods
1274 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1276 Prior to C++11, a common pattern to make a class uncopyable was to declare an
1277 unimplemented copy constructor and copy assignment operator and make them
1278 private. This would give a compiler error for accessing a private method or a
1279 linker error because it wasn't implemented.
1281 With C++11, we can mark methods that won't be implemented with ``= delete``.
1282 This will trigger a much better error message and tell the compiler that the
1283 method will never be implemented. This enables other checks like
1284 ``-Wunused-private-field`` to run correctly on classes that contain these
1287 For compatibility with MSVC, ``LLVM_DELETED_FUNCTION`` should be used which
1288 will expand to ``= delete`` on compilers that support it. These methods should
1289 still be declared private. Example of the uncopyable pattern:
1295 DontCopy(const DontCopy&) LLVM_DELETED_FUNCTION;
1296 DontCopy &operator =(const DontCopy&) LLVM_DELETED_FUNCTION;
1301 Don't evaluate ``end()`` every time through a loop
1302 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1304 Because C++ doesn't have a standard "``foreach``" loop (though it can be
1305 emulated with macros and may be coming in C++'0x) we end up writing a lot of
1306 loops that manually iterate from begin to end on a variety of containers or
1307 through other data structures. One common mistake is to write a loop in this
1312 BasicBlock *BB = ...
1313 for (BasicBlock::iterator I = BB->begin(); I != BB->end(); ++I)
1316 The problem with this construct is that it evaluates "``BB->end()``" every time
1317 through the loop. Instead of writing the loop like this, we strongly prefer
1318 loops to be written so that they evaluate it once before the loop starts. A
1319 convenient way to do this is like so:
1323 BasicBlock *BB = ...
1324 for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I)
1327 The observant may quickly point out that these two loops may have different
1328 semantics: if the container (a basic block in this case) is being mutated, then
1329 "``BB->end()``" may change its value every time through the loop and the second
1330 loop may not in fact be correct. If you actually do depend on this behavior,
1331 please write the loop in the first form and add a comment indicating that you
1332 did it intentionally.
1334 Why do we prefer the second form (when correct)? Writing the loop in the first
1335 form has two problems. First it may be less efficient than evaluating it at the
1336 start of the loop. In this case, the cost is probably minor --- a few extra
1337 loads every time through the loop. However, if the base expression is more
1338 complex, then the cost can rise quickly. I've seen loops where the end
1339 expression was actually something like: "``SomeMap[X]->end()``" and map lookups
1340 really aren't cheap. By writing it in the second form consistently, you
1341 eliminate the issue entirely and don't even have to think about it.
1343 The second (even bigger) issue is that writing the loop in the first form hints
1344 to the reader that the loop is mutating the container (a fact that a comment
1345 would handily confirm!). If you write the loop in the second form, it is
1346 immediately obvious without even looking at the body of the loop that the
1347 container isn't being modified, which makes it easier to read the code and
1348 understand what it does.
1350 While the second form of the loop is a few extra keystrokes, we do strongly
1353 ``#include <iostream>`` is Forbidden
1354 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1356 The use of ``#include <iostream>`` in library files is hereby **forbidden**,
1357 because many common implementations transparently inject a `static constructor`_
1358 into every translation unit that includes it.
1360 Note that using the other stream headers (``<sstream>`` for example) is not
1361 problematic in this regard --- just ``<iostream>``. However, ``raw_ostream``
1362 provides various APIs that are better performing for almost every use than
1363 ``std::ostream`` style APIs.
1367 New code should always use `raw_ostream`_ for writing, or the
1368 ``llvm::MemoryBuffer`` API for reading files.
1375 LLVM includes a lightweight, simple, and efficient stream implementation in
1376 ``llvm/Support/raw_ostream.h``, which provides all of the common features of
1377 ``std::ostream``. All new code should use ``raw_ostream`` instead of
1380 Unlike ``std::ostream``, ``raw_ostream`` is not a template and can be forward
1381 declared as ``class raw_ostream``. Public headers should generally not include
1382 the ``raw_ostream`` header, but use forward declarations and constant references
1383 to ``raw_ostream`` instances.
1388 The ``std::endl`` modifier, when used with ``iostreams`` outputs a newline to
1389 the output stream specified. In addition to doing this, however, it also
1390 flushes the output stream. In other words, these are equivalent:
1394 std::cout << std::endl;
1395 std::cout << '\n' << std::flush;
1397 Most of the time, you probably have no reason to flush the output stream, so
1398 it's better to use a literal ``'\n'``.
1400 Don't use ``inline`` when defining a function in a class definition
1401 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1403 A member function defined in a class definition is implicitly inline, so don't
1404 put the ``inline`` keyword in this case.
1431 This section describes preferred low-level formatting guidelines along with
1432 reasoning on why we prefer them.
1434 Spaces Before Parentheses
1435 ^^^^^^^^^^^^^^^^^^^^^^^^^
1437 We prefer to put a space before an open parenthesis only in control flow
1438 statements, but not in normal function call expressions and function-like
1439 macros. For example, this is good:
1444 for (I = 0; I != 100; ++I) ...
1445 while (LLVMRocks) ...
1448 assert(3 != 4 && "laws of math are failing me");
1450 A = foo(42, 92) + bar(X);
1457 for(I = 0; I != 100; ++I) ...
1458 while(LLVMRocks) ...
1461 assert (3 != 4 && "laws of math are failing me");
1463 A = foo (42, 92) + bar (X);
1465 The reason for doing this is not completely arbitrary. This style makes control
1466 flow operators stand out more, and makes expressions flow better. The function
1467 call operator binds very tightly as a postfix operator. Putting a space after a
1468 function name (as in the last example) makes it appear that the code might bind
1469 the arguments of the left-hand-side of a binary operator with the argument list
1470 of a function and the name of the right side. More specifically, it is easy to
1471 misread the "``A``" example as:
1475 A = foo ((42, 92) + bar) (X);
1477 when skimming through the code. By avoiding a space in a function, we avoid
1478 this misinterpretation.
1483 Hard fast rule: Preincrement (``++X``) may be no slower than postincrement
1484 (``X++``) and could very well be a lot faster than it. Use preincrementation
1487 The semantics of postincrement include making a copy of the value being
1488 incremented, returning it, and then preincrementing the "work value". For
1489 primitive types, this isn't a big deal. But for iterators, it can be a huge
1490 issue (for example, some iterators contains stack and set objects in them...
1491 copying an iterator could invoke the copy ctor's of these as well). In general,
1492 get in the habit of always using preincrement, and you won't have a problem.
1495 Namespace Indentation
1496 ^^^^^^^^^^^^^^^^^^^^^
1498 In general, we strive to reduce indentation wherever possible. This is useful
1499 because we want code to `fit into 80 columns`_ without wrapping horribly, but
1500 also because it makes it easier to understand the code. To facilitate this and
1501 avoid some insanely deep nesting on occasion, don't indent namespaces. If it
1502 helps readability, feel free to add a comment indicating what namespace is
1503 being closed by a ``}``. For example:
1508 namespace knowledge {
1510 /// This class represents things that Smith can have an intimate
1511 /// understanding of and contains the data associated with it.
1515 explicit Grokable() { ... }
1516 virtual ~Grokable() = 0;
1522 } // end namespace knowledge
1523 } // end namespace llvm
1526 Feel free to skip the closing comment when the namespace being closed is
1527 obvious for any reason. For example, the outer-most namespace in a header file
1528 is rarely a source of confusion. But namespaces both anonymous and named in
1529 source files that are being closed half way through the file probably could use
1534 Anonymous Namespaces
1535 ^^^^^^^^^^^^^^^^^^^^
1537 After talking about namespaces in general, you may be wondering about anonymous
1538 namespaces in particular. Anonymous namespaces are a great language feature
1539 that tells the C++ compiler that the contents of the namespace are only visible
1540 within the current translation unit, allowing more aggressive optimization and
1541 eliminating the possibility of symbol name collisions. Anonymous namespaces are
1542 to C++ as "static" is to C functions and global variables. While "``static``"
1543 is available in C++, anonymous namespaces are more general: they can make entire
1544 classes private to a file.
1546 The problem with anonymous namespaces is that they naturally want to encourage
1547 indentation of their body, and they reduce locality of reference: if you see a
1548 random function definition in a C++ file, it is easy to see if it is marked
1549 static, but seeing if it is in an anonymous namespace requires scanning a big
1552 Because of this, we have a simple guideline: make anonymous namespaces as small
1553 as possible, and only use them for class declarations. For example, this is
1563 bool operator<(const char *RHS) const;
1565 } // end anonymous namespace
1567 static void runHelper() {
1571 bool StringSort::operator<(const char *RHS) const {
1585 bool operator<(const char *RHS) const;
1592 bool StringSort::operator<(const char *RHS) const {
1596 } // end anonymous namespace
1598 This is bad specifically because if you're looking at "``runHelper``" in the middle
1599 of a large C++ file, that you have no immediate way to tell if it is local to
1600 the file. When it is marked static explicitly, this is immediately obvious.
1601 Also, there is no reason to enclose the definition of "``operator<``" in the
1602 namespace just because it was declared there.
1607 A lot of these comments and recommendations have been culled from other sources.
1608 Two particularly important books for our work are:
1611 <http://www.amazon.com/Effective-Specific-Addison-Wesley-Professional-Computing/dp/0321334876>`_
1612 by Scott Meyers. Also interesting and useful are "More Effective C++" and
1613 "Effective STL" by the same author.
1615 #. `Large-Scale C++ Software Design
1616 <http://www.amazon.com/Large-Scale-Software-Design-John-Lakos/dp/0201633620/ref=sr_1_1>`_
1619 If you get some free time, and you haven't read them: do so, you might learn